An Algorithmic Approach Based on Data Trees and Genetic Algorithms to Understanding Charged and Neutral Metal Nanocluster Growth

Metal nanocluster growth pathways are challenging to study due to the number of possible species involved and the branching nature of possible mechanisms. We present a data tree-based approach to mapping these reaction pathways based on structures and energies obtained from Density Functional Theory (DFT) computations and including positive, negative, and neutral clusters in a continuum solvent of water. We also develop a genetic algorithm to study the relative stability of the clusters, which we combine with the data tree method to determine the effects that favorable decomposition reactions have on the reaction pathways. Introducing data tree pruning based on the exothermicity of each reaction, and including the first and second most important paths for each cluster up to five atoms including positive, neutral, and negative clusters, is then implemented to determine the cluster growth paths. These most favorable Ag–Ag reaction pathways are in agreement with more limited prior theoretical and experimental results, but they provide more systematic results that include predictions about the importance of clusters not previously identified. A key feature of the data tree approach is that it provides a comprehensive sampling of possible clusters, but without needing to generate the entire reaction tree. Additionally, the data tree-based approach allows for flexibility in the analysis based on restricting reactant charge or size, selecting a starting species to mimic an experimental precursor, and choosing a maximum allowable final cluster size or charge of interest.